Package 'COMBAT'

Title: A Combined Association Test for Genes using Summary Statistics
Description: Genome-wide association studies (GWAS) have been widely used for identifying common variants associated with complex diseases. Due to the small effect sizes of common variants, the power to detect individual risk variants is generally low. Complementary to SNP-level analysis, a variety of gene-based association tests have been proposed. However, the power of existing gene-based tests is often dependent on the underlying genetic models, and it is not known a priori which test is optimal. Here we proposed COMBined Association Test (COMBAT) to incorporate strengths from multiple existing gene-based tests, including VEGAS, GATES and simpleM. Compared to individual tests, COMBAT shows higher overall performance and robustness across a wide range of genetic models. The algorithm behind this method is described in Wang et al (2017) <doi:10.1534/genetics.117.300257>.
Authors: Minghui Wang, Yiyuan Liu, Shizhong Han
Maintainer: Minghui Wang <[email protected]>
License: GPL-2
Version: 0.0.4
Built: 2024-11-04 02:47:08 UTC
Source: https://github.com/mw201608/combat

Help Index

A Combined Gene-based Association Test

Description

This function implements a combined gene-based association test using SNP-level P values and reference genotype data.

Usage

COMBAT(x, snp.ref, vegas.pct = c(0.1,0.2,0.3,0.4,1), pca_cut_perc = 0.995,
   nperm = 100, seed=12345, ncores=1)

Arguments

x

a vector of SNP-level P values.
snp.ref

a matrix of SNP genotypes (coded as allele counts) from reference samples, with samples in rows and SNPs in columns.
vegas.pct

a numeric vector, fraction of the top SNPs to be used in the VEGAS method.
pca_cut_perc

numeric, cutoff for percentage of sum of eigen values in the simpleM approach.
nperm

number of permutations for computing the correlation between P values of different tests.
seed

random seed to derive consistent outcome.
ncores

number of CPU cores for parallel computing.

Details

COMBAT uses simulation and the extended Simes procedure (ext_simes) to combine multiple gene-based association test statistics (currently including gates, vegas, and simpleM) to perform a more powerful association analysis. This method does not require raw genotype or phenotype data, but needs only SNP-level P-values and correlations between SNPs from ancestry-matched samples. The technical details about the method is described in Wang et al (2017) <doi:10.1534/genetics.117.300257>.

Value

A vector of p-values from COMAT and each individual gene-based test.

References

Minghui Wang, Jianfei Huang, Yiyuan Liu, Li Ma, James B. Potash, Shizhong Han. COMBAT: A Combined Association Test for Genes using Summary Statistics. Genetics 2017, 207(3): 883-891. https://doi.org/10.1534/genetics.117.300257.

See Also

ext_simes, gates, vegas, simpleM.

Examples

# read SNP P values
file1 <- paste(path.package("COMBAT"),"extdata","SNP_info.txt.gz",sep="/")
snp.info  <- read.table(file1, header = TRUE, as.is=TRUE)
snp.pvals <- as.matrix(snp.info[,2])

# read reference genotype
file2 <- paste(path.package("COMBAT"),"extdata","SNP_ref.txt.gz",sep="/")
snp.ref   <- read.table(file2, header = TRUE)
snp.ref   <- as.matrix(snp.ref)
#call COMBAT

COMBAT(snp.pvals, snp.ref, nperm=100, ncores=2)

Extended Simes Procedure

Description

Combine a vector of test P values by correction for number of independent tests.

Usage

ext_simes(x, cor_r)

Arguments

x

a vector of SNP-level P values.
cor_r

correlation among P values.

Value

P value.

See Also

COMBAT

Examples

# see ?COMBAT

Calling Gene-based Association Tests

Description

Several gene-based association tests methods are implemented.

Usage

gates(x, cor_G)
vegas(x, cor_G, vegas.pct=c(0.1,0.2,0.3,0.4,1), max.simulation=1e6)
simpleM(x, cor_G, pca_cut_perc=0.995)

Arguments

x

a vector of SNP-level P values.
cor_G

SNP-SNP correlation matrix.
vegas.pct

a numeric vector, specifying the fraction of the top SNPs to be used in the VEGAS method.
max.simulation

maximum number of simulations to be performed. Must be at least 1e6.
pca_cut_perc

cutoff for percentage of sum of eigen values.

Details

Function gates implements the GATES method (Li et al 2011, American Journal of Human Genetics 88:283-293), vegas implements VEGAS with different proportion tests (Liu et al 2010, American Journal of Human Genetics 87:139-145), and simpleM is the simpleM method (Gao et al 2008, Genetic Epidemiology 32:361-369).

Value

P value(s).

See Also

COMBAT, ld.Rsquare.

Examples

# read SNP P values
file1 <- paste(path.package("COMBAT"),"extdata","SNP_info.txt.gz",sep="/")
snp.info  <- read.table(file1, header = TRUE, as.is=TRUE)
snp.pvals <- as.matrix(snp.info[,2])

# read reference genotype
file2 <- paste(path.package("COMBAT"),"extdata","SNP_ref.txt.gz",sep="/")
snp.ref   <- read.table(file2, header = TRUE)
snp.ref   <- as.matrix(snp.ref)

#compute correlation among SNPs
cor_G <- ld.Rsquare(snp.ref)

#call gates
(pval_gates <- gates(x=snp.pvals, cor_G=cor_G))

#call vegas
(pval_vegas <- vegas(x=snp.pvals, cor_G=cor_G))

#call simpleM
(pval_simpleM <- simpleM(x=snp.pvals, cor_G=cor_G))

Linkage Disequilibrium Estimate as Correlation

Description

Compute linkage disequilibrium among SNPs using correlation coefficients.

Usage

ld.Rsquare(x)

Arguments

x

a matrix of SNP genotypes with samples in the rows.

Value

A positive definite correlation matrix.

See Also

gates, vegas, simpleM.